Tip
How to not lose your prompt work
A four-prompt workflow for anyone iterating on a prompt across multiple sessions — so you always know where you are and can hand it off cleanly.
Here's what happens when you build a prompt through conversation: the decisions are scattered across the chat. You tweak the tone, adjust the criteria, add an exception for the edge case you discovered — and by the next session, you're not sure what the current version actually is, or why you made the choices you made.
This workflow is four prompts. Run them at the right moments and the session stays coherent, the handoff is clean, and you're not starting from scratch every time.
💡 Works with any AI tool
These prompts work in Claude, ChatGPT, Copilot, Gemini — anywhere you're having a multi-turn conversation to build or refine something. The tool doesn't matter. The pattern does.
The workflow
1
Get the clean current version
During session
Run this whenever you make meaningful changes — not just at the end.
Based on everything we've worked on in this session, please give me the complete current version of the prompt as it stands right now — the full text, ready to copy and use, with all the changes and additions we've made incorporated. Don't summarize it or explain it, just give me the clean complete version.
Copy prompt
This is your source of truth. Copy it somewhere you'll find it — a note, a doc, a folder — so you always have the current working version without reconstructing it from memory.
2
Continuity summary
End of session
Run this before you close out so the next session picks up where this one left off.
We've been working on a prompt together in this session. Before I close out, I need a continuity summary so I can pick up where we left off next time. Please summarize: what the prompt is designed to do, what specific decisions we made today and why we made them, what's working well, what still needs work or testing, and what I should try or think about in the next session. Keep it concise enough that I can paste it at the start of my next conversation to get back up to speed quickly.
Copy prompt
Paste this summary at the very start of your next session before you do anything else. The AI will have the full context and you won't have to re-explain where you are.
3
Plain language explanation
When prompt is done
Run this when the prompt feels finished and ready to share or hand off.
I've built a prompt with your help that I need to hand off to other people to use. Please explain what this prompt does in plain language — what it's designed to do, what logic or criteria it's applying, what good input looks like, what the output means, and what someone should do if the output doesn't look right. Write it as documentation for someone who didn't build this and won't have me there to explain it.
Copy prompt
This becomes the documentation that travels with the prompt. Anyone who uses it should be able to read this and understand what it does — without asking you.
4
Logic check
Before handoff
Run this after the plain language explanation. Read it carefully before sharing.
Walk me through the logic this prompt is using, step by step — specifically what it's checking for, how it's making its decisions, and what it's producing. I want to verify it's doing what I intended before other people use it.
Copy prompt
If the AI describes logic you didn't intend, or explains something you don't recognize — that's your flag. Fix it before anyone else uses it. This is the quality check that people skip and then regret.
🤝 The step most people skip
Prompts 3 and 4 feel like extra work when you're done and just want to share the thing. They're not. They're the difference between "here's something I built" and "here's something that works the way I think it does." Run both.
Quick reference — when to run each one
1
Clean current versionAfter any meaningful change
2
Continuity summaryEnd of every session
3
Plain language explanationWhen the prompt is done
4
Logic checkBefore handoff to anyone else